Loading…
Acquisition of endometrium ultrasound scans and its effect on the results of computer analysis of texture
Introduction: The computer analysis of ultrasound scans becomes more and more common in diagnostic procedure in many disorders. Objectives: The aim of the study was to evaluate how the parameters of the image acquisition (ultrasound penetration depth) determine results of computer analysis of textur...
Saved in:
Published in: | Przegla̜d menopauzalny 2009-05, Vol.13 (3), p.127 |
---|---|
Main Authors: | , , , , |
Format: | Article |
Language: | eng ; pol |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Introduction: The computer analysis of ultrasound scans becomes more and more common in diagnostic procedure in many disorders. Objectives: The aim of the study was to evaluate how the parameters of the image acquisition (ultrasound penetration depth) determine results of computer analysis of texture. Material and methods: Menopausal women with abnormal uterine bleeding were subjected to the study. Ultrasoud examination was performed before D&C procedure. Endometrial ultrasound scans were recorded for 4 different ultrasound penetration depths (6 cm, 8 cm, 10 cm and 13 cm). Two groups of patients were analysed. First group included scans of 6 patients suffering from endometrial cancer, second group scans of patients with other endometrial disorders and normal endometrium (degeneratio cystica - 3 patients, endometrium dyshormonoticum - 3 patients, polypus endometrialis - 3 patients, endometrium proliferans - 3 patients, endometrium secretans - 3 patients). Conclusions: Results of the analysis indicates that the ultrasound penetration depth affects values of textural features. In cases of endometrial examination the best way to discriminate images of the patients with confirmed adenocarcinoma from the images with other disorders was observed when the ultrasound penetration depth of 6 cm and 8 cm was applied. The most discriminant texture features are derived from wavelet transform and autoregressive model. The best feature selection method is Fisher coefficient based approach. |
---|---|
ISSN: | 1643-8876 2299-0038 |